security-scanner_skill

This skill helps you audit installed plugins and skills for security risks using AI analysis before installation.

20

GitHub Stars

1

Bundled Files

3 weeks ago

Catalog Refreshed

2 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstart where the catalogue uses aiagentskills.

npx veilstart add skill hiroro-work/claude-plugins --skill security-scanner

SKILL.md18.5 KB

Overview

This skill scans installed plugins and skills for security risks, including malicious code and malicious natural language instructions. Use the /security-scanner command to audit before installation or to inspect GitHub repositories. It detects both dangerous code patterns and deceptive or exfiltration instructions in documentation and prompts.

How this skill works

The scanner locates plugin and skill directories at user and project levels (and supports multiple agent paths) or fetches content from public GitHub URLs. It reads metadata and all relevant files, then applies semantic AI analysis and pattern checks to detect code-level threats and malicious natural-language instructions. It reports findings, skips trusted sources, and can operate in a strict --all mode that ignores trusted exclusions.

When to use it

Before installing a third-party plugin or skill
When auditing local skills/plugins after pulling new code
When reviewing a public GitHub plugin or skill before use
As part of a security review for multi-agent setups
To verify changes after merging new skill updates

Best practices

Run scans on both project-level and user-level paths to avoid blind spots
Maintain a local trusted_sources config to reduce noise for known good packages
Use --url to scan public GitHub repos before pulling code locally
Use --all only when you want a full audit that ignores trusted exclusions
Review both code results and natural-language findings; treat language-based risks as high priority

Example use cases

Audit a downloaded Claude Code plugin in ~/.claude/plugins before enabling it
Scan a project’s .claude/.codex skills directories during CI before deployment
Fetch and analyze a skill directory from a public GitHub URL to confirm safety
Run a user-level scan to detect malicious instructions embedded in agent system prompts
Use --project when evaluating new skills added to a repository branch

FAQ

Only public GitHub repositories are supported. Other hosts return an unsupported-host error.

Can I skip known trusted plugins or marketplaces?

Yes. Configure trusted_marketplaces, trusted_plugins, and trusted_skills in a local settings file; project-level settings override user-level.

Does it detect malicious natural language instructions?

Yes. The scanner analyzes system prompts, skill instructions, and documentation for data-exfiltration, deception, and security-bypass instructions and flags them as critical.